基于粒子群优化支持向量机的电梯故障诊断

电梯故障时，具有故障特征提取困难和故障类型识别率低的问题。因此，拟提取其振动信号并进行分析，找到故障特征。然而，鉴于其振动信号为非平稳、非高斯且背景噪声较大的信号，给有效辨识造成很大困难，所以，提出应用最优小波包分解和最小二乘支持向量机相结合进行电梯智能故障诊断的方法。借助最优小波包理论，首先提取电梯故障振动信号的能量分布；然后将其能量分布与时域指标相结合，构造故障特征向量；最后，将故障特征向量作为粒子群算法优化最小二乘支持向量机的输入对电梯故障类型进行识别。仿真结果表明，最优小波包理论与最小二乘支持向量机相结合的故障诊断技术发挥了两者的优势，证明了该方法的有效性和实用性。

Diagnosis of Elevator Faults with SVM Based on Optimization by PSO

In case of elevator faults, there will be the problems including fault feature extraction difficulties and low rate of fault type identification. However, considering that the vibration signals are non-stationary and non-Gaussian signals with large background noise, optimal wavelet packet decomposition was combined with least squares support vector machine to diagnose intelligent elevator faults. Firstly, the energy distribution of elevator fault vibration signal is extracted by dint of optimal wavelet packet theory. Then, its energy distribution is combined with time-domain indices to constitute fault feature vectors. Finally, fault feature vectors are used to identify elevator fault types as the input of the least squares support vector machine of particle swarm optimization. As shown by the simulated result, the failure diagnosis technique, which integrates optimal wavelet packet theory with least squares support vector machine, has given full play to their respective advantages, which proves that this method is effective and practical.